Summary
According to a post on X by @rohanpaul_ai, a Stanford paper tested 10 frontier AI models in trading, bargaining and auction tasks, changing only how precisely each side's private values were shared. Full information lowered Claude Opus 5.5's share of possible gains from 94% to 64% in one auction setting, and cut Grok 4.7 from 97% to 62%. In trade, however, Claude Sonnet 5 rose from 72% to 86%. The post's headline frames the result as more information making some agents greedier.
AI-assisted summary based on the listed source.
What happened
A Stanford paper ran 10 frontier models in trading, bargaining and auctions and changed only how precisely each side’s private values were shared. Full information cut Claude Opus 5.5’s share of possible gains from 94% to 64% in one auction setting and dropped Grok 4.7 from 97% to 62%, but raised Claude Sonnet 5...
Why it matters
Companies that let AI agents negotiate on their behalf may see different outcomes depending on how much information each side's agent has about the other's private values. The reported results suggest that sharing more information does not help every model in every setting, so outcomes may depend on both the model and the task. Watch for the full paper to see how the tests were designed and whether the pattern holds beyond the examples in the post.
Signal Intelligence
Signal Strength 95%
Technical label WATCH
Public Interest 39
Category RESEARCH
Reader Depth TECHNICAL
Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.
Public Interest components
Recognizable Entity Score 60
Practical Impact Score 0
Novelty Interest Score 70
Consequence Score 18
Curiosity Score 16
Shareability Score 54
Why this is here
VQV surfaced this signal because it is recent, relevant to Consumer AI, connected to Scobleizer News Pulse (X).
